Post on 15-Apr-2018
transcript
CONSTRAINING COSMOLOGICAL PARAMETERS AND
TESTING THE ISOTROPY OF THE HUBBLE
EXPANSION USING SUPERNOVAE Ia
Bachelor of Science Thesis
By
Konstantinos Nikolaos Migkas 1
Thesis Supervisor:
Prof. Manolis Plionis
1Physics Department, Aristotle University of Thessaloniki, Thessaloniki
Greece 54124
February, 2015
1
Abstract
In this thesis we use the latest Union2.1 Supernovae Ia sample in order to put constraints
to the basic cosmological parameters, such as the Dark Energy equation of state parameter
w , the matter density Ωm and the deceleration parameter q0. We further use the angular
coordinates of the SNe Ia in order to divide the sample in various independent groups,
such as different solid angles, and test the isotropy of the Hubble expansion. Our main
results are:
1. We have reproduced all the published results regarding the cosmological constrains
put by the SNe Ia Hubble expansion probe.
2. We have estimated the deceleration parameter q0 using the low-z SNe Ia (z ≤ 0.24)
and found q0 ' −0.501 corresponding, for a flat Universe, to a matter density
Ωm ' 0.333.
3. We have found that the Ωm-w contours, provided by the SNe Ia analysis of different
redshift ranges, cover distinct regions of the Ωm-w solution space. We have also
found, interestingly, that excluding the intermediate redshift range (0.314 < z ≤
0.57) does not affect significantly the cosmological constraints in the Ωm -w plane,
a fact that can have important implications for future observation strategies.
4. We have identified one sky region, with galactic coordinates 35o < l < 83o &
−79o < b < −37o, containing 82 SNe Ia (15% of total with z ≥ 0.02), that shares a
different Hubble expansion behaviour indicating a possible anisotropy, if confirmed.
We have excluded as a possible cause systematic effects related to the different
surveys that constitute the Union2.1 set. An alternative explanation, that we will
investigate in the future is the existence of a bulk flow that could affect significantly
the redshifts of the specific SNe Ia.
3
Περίληψη
Σε αυτή την εργασία χρησιμοποιούμε το πιο πρόσφατο σετ Υπερκαινοφανών Τύπου Ια
Union2.1 προκειμένου να υπολογίσουμε τα όρια τιμών των βασικών κοσμολογικών παραμέτρων,
όπως ο βαροτροπικός δείκτης w της Σκοτεινής Ενέργειας, η πυκνότητα ύλης Ωm και ο
παράγοντας επιβράδυνσης q0. Επίσης, χρησιμοποιούμε τις γαλαξιακές συντεταγμένες των
Υπερκαινοφανών Ια για να χωρίσουμε το δείγμα σε διάφορες ανεξάρτητες ομάδες, όπως π.χ.
σε στερεές γωνίες, και να ελέγξουμε την ισοτροπία της διαστολής Hubble. Τα βασικά μας
αποτελέσματα είναι:
1. Αναπαράγαμε όλα τα δημοσιευμένα αποτελέσματα που αφορούν τα όρια των κοσ-
μολογικών παραμέτρων που προκύπτουν από τους Υπερκαινοφανείς Ια.
2. Υπολογίσαμε τον συντελεστή επιβράσυνσης q0 χρησιμοποιώντας τους Υπερκαινο-
φανείς Ια με χαμηλές ερυθρομεταθέσεις (z ≤ 0.24) και βρήκαμε q0 ' −0.501, που
αντιστοιχεί, για ένα επίπεδης γεωμετρίας Σύμπαν, σε μια πυκνότητα ύλης Ωm ' 0.333.
3. Βρήκαμε ότι οι ισοπίθανες επιφάνειες των Ωm -w που προκύπτουν από την ανάλυση
των Υπερκαινοφανών Ια για διαφορετικά όρια ερυθρομεταθέσεων, καλύπτουν ανόμοιες
περιοχές στην περιοχή λύσεων των Ωm -w . Επίσης, με πολύ ενδιαφέρον, βρήκαμε ότι
αν αποκλείσουμε τις μεσαίες ερυθρομεταθέσεις (0.314 < z ≤ 0.57, 1/4 του συνο-
λικού δείγματος) δεν επηρεάζονται σημαντικά τα αποτελέσματα και τα όρια των κοσ-
μολογικών παραμέτρων.
4. Ανιχνεύσαμε μία περιοχή του ουρανού με γαλαξιακές συντεταγμένες 35o < l < 83o ·
−79o < b < −37o στην οποία περιέχονται 82 Υπερκαινοφανείς Ια (το 15% του
συνολικού δείγματος με z ≥ 0.02) η οποία χαρακτηρίζεται από μια διαφορετική δι-
αφορά στην διαστολή υποδεικνύοντας μια πιθανή ανισοτροπία στη διαστολή, αν τελικά
επιβεβαιωθεί. Αποκλε;ισαμε ως πιθαν;η αιτ;ια ;ενα συστηματικ;ο σφ;αλμα που σχετ;ιζεται
με διαφορετικο;υς καταλ;ογους δεδομ;ενων τα οπο;ια απαρτ;ιζουν το δε;ιγμα Union2.1 .
Μια εναλλακτική εξήγηση που θα εξετάσουμε στο μέλλον, είναι η ύπαρξη μιας συνεκ-
τικής ροής που θα επηρέαζε τις ερυθρομεταθέσεις των δεδομένων.
4
Acknowledgements
I would like to express my deep gratitude to Prof. Plionis Manolis for guiding and helping
me throughout this thesis. He was always more than willing to spent time with me in
order to solve my queries. His deep knowledge in the field of observational cosmology and
his ideas for this project played a key role in completing it. I am also thankful to Maria
Manolopoulou who provided me with her valuable help about programming tools at the
beginning of my work. Finally, I owe a lot to my lovely girlfriend and fellow student
Eftychia Madika for all her constructive advice, for her daily support and for believing
so much in me.
5
Contents
1 Introduction 8
1.1 Basics of Dynamical Cosmology . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 Taking advantage of standard candles . . . . . . . . . . . . . . . . . . . . 10
1.3 How SNe Ia are produced? . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 The light curves of SNe Ia . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Cosmic and galactic dust . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.6 Other cosmological probes . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2 Theoretical expectations 21
2.1 Different QDE and ΛCDM models comparison . . . . . . . . . . . . . . . 21
2.2 Models comparison for a non-flat Universe (Ωk 6= 0) . . . . . . . . . . . . 23
2.3 Model comparison for CPL parametrization (w = w(z)) . . . . . . . . . . 23
3 The SNe Ia data 26
3.1 Criteria of confirmation of SNe Ia type and quality cuts . . . . . . . . . . 26
3.2 Redshift distribution of the Union2.1 sample . . . . . . . . . . . . . . . . 27
3.3 Mapping the Union2.1 sample . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4 Statistical uncertainties of the distance moduli . . . . . . . . . . . . . . . 32
4 Cosmological parameters fitting 34
4.1 One-parameter models (Ωm or w) . . . . . . . . . . . . . . . . . . . . . . 34
4.2 One-parameter model (q0) . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Two-parameter model (Ωm and w) . . . . . . . . . . . . . . . . . . . . . 37
4.4 Hubble flow divided in different redshift bins . . . . . . . . . . . . . . . . 40
5 Investigating possible Hubble expansion anisotropies 45
5.1 Among the two galactic hemispheres . . . . . . . . . . . . . . . . . . . . 45
5.2 Among random groups of SNe Ia . . . . . . . . . . . . . . . . . . . . . . 48
6 Joint likelihood analysis 55
6.1 Joint likelihood analysis for the two galactic hemispheres . . . . . . . . . 55
6
6.2 Joint likelihood analysis for the three redshift bins . . . . . . . . . . . . . 56
6.3 Joint likelihood analysis of ten independent SNe Ia subsamples . . . . . . 57
7 Further analysis for the unusual sky region 59
8 Conclusions 65
Appendices 66
A χ2 minimization, our data analysis method 66
B Joint likelihood analysis 68
9 References 69
7
1 Introduction
It is well-known that we live in the Golden Age of Cosmology. The amount of observa-
tional data available is higher than ever. The detailed observations of the cosmos in the
20th century has changed dramatically our ideas for it. Millions of galaxies were discov-
ered, we realized that the Universe is expanding, we found that there are huge amount
of Cold Dark Matter not seen whose nature we still do not know and we predicted and
confirmed the existence of Cosmic Microwave Background (CMB) radiation. However,
maybe the biggest breakthrough of all, was the discovery by Saul Perlmutter, Adam Riess
and Brian P. Schmidt in 1998 that the expansion of the Universe is accelerating, some-
thing completely contrary to what was believed until then. Ever since, the cosmology
community has devoted a large amount of research in order to understand the cause for
the accelerated expansion as well as the nature of the Dark Matter.
1.1 Basics of Dynamical Cosmology
One of the main tools we use to study the dynamical behaviour of the Universe are
the Friedmann equations. Alexander Friedmann in 1922, achieved, using Einstein’s field
equations of gravity to derive a set of equations that described the dynamical behaviour of
the Universe. The content of the Universe is assumed to be a homogeneous perfect fluid,
that is a very good approximation to reality, according to the Cosmological Principle and
the Robertson-Walker metric, which assumes homogeneity and isotropy. Therefore, these
equations describe homogeneous and isotropic FLRW1 models and they are:
H2 =8πGρ
3− kc2
α2+
Λc2
3
H +H2 = −4πG
3
(ρ+
3P
c2
)+
Λc2
3
(1.1)
where H ≡ α
αis the Hubble parameter, with α being the scale factor of the Universe, G
1Friedmann-Lemaıtre-Robertson-Walker
8
the gravitational constant, c the speed of light, ρ the mass density of the total cosmic
(matter and radiation) fluid, Λ the cosmological constant and k = 0,±1 is depended of
the spatial curvature of the Universe. We have k = 0 for a spatially flat Universe and
k = +1,−1 for a positively or negatively curved Universe respectively. Using eq. (1.1)
we can have:
1 =8πGρ
3H2− kc2
α2H2+
Λc2
3H2
=8πG
3H2[ρ+ ρk + ρΛ] ⇒
1 = Ωm + Ωk + ΩΛ
(1.2)
where Ωm, Ωk, ΩΛ are the fractional densities of matter, curvature and the cosmolog-
ical constant or Dark Energy. Dark Energy (hereafter DE) is the term we use for this
hypothetical fluid (with a negative equation of state parameter) that is expressed by ΩΛ
and seems to dominate the Universe in terms of mass-energy density. From eq. (1.1) we
can find the total density (of all source terms mass, radiation, cosmological constant and
curvature), and this is
ρtot =3H2
8πG≈ 10−29 gr/cm3
for this era. Thus, each density parameter is Ωi =ρiρtot
. Combining the two parts of eq.
(1.1) we can derive the continuity equation, that describes the changes of mass-energy
density over time:
ρ+ 3α
α
(ρ+
P
c2
)= 0 ⇒
ρ = −3H
(ρ+
P
c2
)⇒
ρ = −3Hρ(1 + w)
(1.3)
where P = wc2ρ, the equation of state of a perfect fluid and w the equation of state
9
parameter. Finally, we can define the deceleration parameter
q = − ααα2
= −
(1 +
H
H2
)(1.4)
Since 1998, thanks to the work of S. Perlmutter, A. Riess and B. P. Schmidt we know
that q0 < 0, which means that the expansion of the Universe is accelerating, and not
decelerating as would happen if our Universe contained only matter, baryonic or not.
For this to happen, it must be H > 0 or in other words Λ > 0. Generally, the second
derivative of the cosmological scale factor must be α > 0. We have to assume that there
is some form of a perfect fluid, DE, with w < −1/3 in order to have an accelerating
expansion. For baryonic and non-baryonic matter we have w = 0 and for the curvature,
w = −1/3.
From the Friedmann and continuity equations, we can easily derive an equation that
describes how the value of the Hubble parameter depends on the density parameters, the
DE equation of state parameter and redshift z (or time):
H(z) = H0
√Ωm(1 + z)3 + Ωk(1 + z)2 + ΩΛ exp
(3
∫ z
0
1 + w(x)
1 + xdx
)(1.5)
where H0 = 70 km/s/Mpc is the value of the Hubble parameter at this time. The
prevailing values of all these parameters according to the latest data are Ωm ≈ 0.3,
Ωk ≈ 0, ΩΛ ≈ 0.7 and w ≈ −1.
For example, for these exact values, when z = 1 (or 7.7×109 years ago) it was H(z = 1) =
123.2 km/s/Mpc while for a matter-dominant Einstein-de Sitter Universe, it would be
H(z) = 198 km/s/Mpc. Finally, as shown in eq. (1.1), (1.3) and (1.5), for a Λ-dominated
Universe with w = −1, the Hubble parameter would be constant throughout time.
1.2 Taking advantage of standard candles
We characterize as standard candles some astrophysical objects which have a known
constant luminosity which is independent of spatial position and time. This gives as
10
0
100
200
300
400
500
600
700
800
900
1000
1100
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
H(z)
z
w=-1, Ωm=0.3,ΩΛ=0.7
w=0, Ωm=1,ΩΛ=0
w=-1, Ωm=0,ΩΛ=1
Figure 1: . The Hubble parameter as evolves with the redshift for Einstein-de Sitter model
(green dashed line), ΛCDM model (red line) and de Sitter model (blue dashed line).
the huge advantage of knowing the absolute magnitude M of the object. The most
common standard candles are the Cepheid variables and the Supernovae of Type Ia (SNe
Ia). Cepheid variables are used to measure the distances of their host galaxies up to
30 Mpc and to calculate the Hubble constant H0. On the other hand, SNe Ia are used for
cosmological reasons and cosmological distances. If we measure the apparent magnitude
m of a standard candle, then knowing its luminosity means that we know its distance
modulus µ, defined as:
µ = m−M = 5 log dL + 25 (1.6)
where dL is the luminosity distance of the object in Mpc. It is obvious that for standard
candles, since we knowM and we can easily measurem, we can calculate the observational
distance modulus of the object µobs while we also measure the object’s redshift zOn the
other hand, we can theoretically calculate the luminosity distance dL and of course the
distance modulus µth for a given model. For a Universe with a non-flat geometry (Ωk 6= 0),
dL is defined as:
11
dL =c(1 + z)√|Ωk|
sinh
[√|Ωk|
∫ z
0
dx
H(x)
](1.7)
while for a flat Universe (Ωk = 0) is:
dL = c(1 + z)
∫ z
0
dx
H(x)(1.8)
Luminosity distance depends strongly on the cosmological model that we use, since the
Hubble parameter within the integral contains all the cosmological parameters that we
are interested in. Using various methods we can estimate the values of the cosmolog-
ical parameters for which the theoretical expected µth gets as close as possible to the
observational µobs. We analyse the method we use in this thesis, the reduced χ2, in the
Appendix.
1.3 How SNe Ia are produced?
The mechanism that leads to a supernova explosion of type Ia, is thought to be the same,
independent of the time and location of the Universe that it occurs. SNe Ia appear in
binary star systems. When one of the stars consumes all its fuel and thermonuclear fusing
is not possible anymore, the star collapse under its own gravity since thermal pressure
cannot support the outer layers. It passes from several stages of stellar evolution (i.e. red
subgiant, giant etc.) that depend on the star’s original mass. For stars that their mass,
when they enter the main sequence is, M ≤ 5 M their final state is that of an white
dwarf. These are almost the 90% of all known stars.
White dwarfs are the product of the deterrence of the gravitational collapse of a star with
ordinary mass by the pressure of degenerate electrons of its core (Varvoglis, Seiradakis
1994). The interior of the white dwarf mainly consists either of helium (He) for those
with an original star mass of M ≤ 3 M, or of a mix of carbon and oxygen (C-O) for
those with an original star mass of 3 M < M ≤ 5 M. The typical values of mass,
density, radius and temperature of a white dwarf are M ∼ 0.7 M, ρ ∼ 106 gr/cm3,
12
R ∼ 7000 km and T ∼ 106 − 107K respectively. Since the gravitational pressure at the
center of the white dwarf is approximately given by
Pc ≈3
8π
GM2
R4(1.9)
then for M = 1 M and R = 6000 km we obtain Pc ≈ 7.7 × 1022 dyn/cm2. Matter in
the interior of white dwarfs is fully ionized.
Due to the Pauli exclusion principle that states that two identical fermions cannot occupy
the same quantum state at the same time and Heisenberg’s uncertainty principle, that
states ∆x ·∆px ≥ h, with ∆x and ∆px the uncertainties of the position and momentum of
the free particles respectively and h is the Planck constant, the free electrons in the interior
of the white dwarf have a non-zero kinetic pressure. This degenerate electron pressure is
larger than the thermal pressure for more than an order of magnitude. Since the radius
of a white dwarf is inversely proportional to its mass, for high-mass white dwarfs (more
electrons, less space) the space that corresponds to each electron ∆x decreases, so ∆p
and electrons velocity increase. Of course the velocity of the electrons cannot be larger
than the speed of light, so it has an upper limit. For relativistic electrons (high-mass
white dwarfs) the degenerate electron pressure is
Pe =1
8
(3
π
)1/3
hc
(ρ
µemp
)4/3
(1.10)
where mp is the proton mass, µe = A/Z ≈ 2 is the mean molecular weight for a carbon-
oxygen white dwarf, ρ the average density and c the speed of light. For the same values
of mass and radius as before we obtain ρ = 2.2 × 1022 gr/cm3 and thus the degenerate
electron pressure is Pe ≈ 1023 dyn/cm2. As we can see this quantum electron pressure is
capable of resisting the gravitational collapse.
There is a limit of mass though that gravitational pressure becomes so powerful that
degenerate electron pressure is unable to resist it. This mass limit is called Chandrasekhar
limit. Its value can be found by equating the gravitational pressure with the quantum
pressure. For Pc = Pe we obtain a mass limit, independent of the radius R, which is
13
Mch ≈ 1.44 M. A white dwarf cannot have M ≥ Mch. Of course this value of Mch
applies for slow rotating white dwarfs. If someone considers a rapidly rotating white
dwarf, then its mass can exceed this limit, Mch. For a fast rotating C-O white dwarf Mch
is calculated to be 1.8M.
Therefore, in a binary stars system, in which one of the stars becomes a C-O white dwarf,
due to its high density (about 1 million times larger than the Sun’s density), it has a
tremendous gravity. At the surface of a typical white dwarf, we have gWD ≈ 104 gSun,
where gSun is the surface gravity of Sun. Due to the white dwarf’s strong gravitational
field, it can gradually accrete mass from its companion star, which is usually thought to
be a main sequence star, a red giant or a helium star (Yoon and Langer 2004), increasing
its total mass. Since the rate of the accretion is usually M ≥ 10−7 M/yr, this process
can last for about 1 million years. As a result, the C-O white dwarf can finally reach
the critical mass of Mch. The corresponding central density is then ρ ∼ 109 gr/cm3. At
this point, degenerate electron pressure can no longer resist the gravitational collapse.
The white dwarf’s radius decreases rapidly, causing the central pressure to increase and
thermonuclear fusion of burning carbon and oxygen take place in its core. This energy-
producing reactions are:
126 C +12
6 C →2010 Ne+4
2 He+ 4.6 MeV
168 O +16
8 O →2814 Si+4
2 He+ 9.6 MeV
Carbon is the main source of energy since it burns completely, unlike oxygen. The total
energy that is released by these reactions is E ∼ 1044 J , and as a result the star explodes.
This explosion is what we call a Type Ia Supernova (SN Ia). As expected, this large
amount of energy leads to a rapid increase of the luminosity L and correspondingly to
the absolute magnitude M . Since the production mechanism for SNe Ia is always the
same, the peak luminosity of all SNe Ia is considered to be the same, not depending
on time (redshift) and place (coordinates). This is the reason why they are considered
standard candles. Their typical values are L ≈ 5 · 109 L and M ≈ −19.3 ± 0.03 in the
B-band. In many cases, a SN Ia may be even more luminous than its host galaxy.
14
1.4 The light curves of SNe Ia
The luminosity of each SN Ia evolves with time. The maximum luminosity appears a few
dozen days after the explosion and then starts to drop until it fades completely in about
a year. The light curves of many SNe Ia have deviations in terms of maximum absolute
magnitudes. The faster the SN Ia luminosity drops with time, the lower the maximum
luminosity. Thus, the SNe Ia with the most negative absolute magnitudes are fading with
a lower rate than the others. This is usually called the time-scale ”stretch-factor” (Kim,
et al. 1997).
Figure 2: Absolute magnitude MB − 5 log (h/65) where h = H0/100 in the B-band versus
time from the peak luminosity. Upper panel: As measured. Bottom panel: After the correction
for the timescale ”stretch-factor”.
15
In the upper panel of Figure 2 we can see the differences among various SNe Ia light
curves. The vast majority of the light curves lie on the average light curve (the yellow
one). After we stretch (or contract) the time scales of the other light curves and adjust
the peak absolute magnitudes, we get the bottom panel of Figure 2, where all light curves
coincide (Perlmutter et al. 1999).
In order to calculate exactly the correct absolute magnitude, someone have to know the
distance of several SNe Ia by other methods, with great accuracy and measure their max-
imum apparent magnitudes. Then, from eq. (1.6) is easy to find the absolute magnitude
M .
1.5 Cosmic and galactic dust
Interstellar dust is a very important component of the Universe, as it exists in abundance
mostly in spiral galaxies (galactic dust), but there is also a lesser amount of gray dust in
the intergalactic medium (cosmic dust). This dust interacts with the light emitted from
the SNe Ia and causes a reddening of the light we receive since the extinction in the blue
light is much larger. This extinction also causes the dimming of the SNe Ia and, if not
corrected, it can lead to the luminosity distance to be overestimated compared with that
of a Universe that contains no dust (Corasaniti 2006).
After the first indications for an accelerating expansion of the Universe in 1998, the
extinction due to the cosmic gray dust was proposed as the reason for the observing
dimness of the SNe Ia (Aguirre 1999a). This idea has been ruled out in the few next
years for several reasons. First of all, for the Universe to be an Einstein-de Sitter Universe
with Ωm = 1, ΩΛ = 0, it would need to have an intergalactic grey dust density of ΩIGD ≈
5× 10−5 in order to explain the 0.5 mag deviation at z = 0.7 as being due to extinction.
Such a density though, according to Riess et al. (1998), would cause redundant reddening
and light dispersion in the SNe Ia and this is contrary to the observations. In addition,
Perlmutter et al. (1999) found that the intrinsic dispersions at z ≈ 0.05 (from Hamuy et
al. 1996) and z ≈ 0.5 (from Supernova Cosmology Project) were very similar (0.154 ±
0.04 and 0.157 ± 0.025 respectively) and this indicates that the processes that lead to
16
this dispersion are not affected by the redshift (Aguirre 1999b). Moreover, Perlmutter
et al. (1998) studied the mean color difference between low-z and high-z data and
found that they were identical, with |E(B − V )|Hamuy = 0.033 ± 0.014 and |E(B −
V )|SCP = 0.035± 0.022, which shows that there is no significant correlation between the
reddening and the redshift of the SNe Ia. As more SNe Ia discovered with time, the idea
of intergalactic, cosmic dust being responsible for the deviation of the observations from
a matter-dominated Universe was ruled out.
However, intergalactic dust does affect the incoming light of the SNe Ia to a certain
degree. Cosmic dust can consists of grains of several materials and sizes, with lots of
studies (Shustov & Vibe 1995; Davies et al. 1998; Bianchi & Ferrara 2005) suggesting that
these sizes are between 0.02 µm−0.2 µm . Since the extinction due to dust is AB ∼ 1/α,
larger grains mean lesser extinction. Bianchi & Ferrara came to the conclusion that the
distribution of the grains probably remains nearly flat (BF model) while the MRN model
(Mathis, Rumpl, Nordsiek) describes the distribution of these grains as N(α) ∝ α−3.5.
Corasaniti (2006) as well as Menard et al. (2009) and various other studies is find that
the extinction caused by intergalactic dust is about AB ≈ 0.01 mag at z = 0.5 for
graphite and even smaller for silicate. At larger redshifts (z ≈ 1.5) for the BF model we
have AB ≈ 0.08 for graphite at and AB ≈ 0.05 for silicate, while for the MRN model
extinction, it is about 0.02 mag larger for every material. In any case, the deviation in
magnitude is within the uncertainties of the observations. If this extinction is not taken
into account it leads the confidence regions of Dark Energy equation of state parameter
w to more negative values (Corasaniti 2006) and the matter density Ωm to lower values.
If we consider this phenomenon, then the real apparent magnitude mreal that we have to
use in eq. 1.6 is mreal = mobs − AB. So, we calculate a smaller luminosity distance dL
and the values of the constrained cosmological parameters change. Menard et al. (2010)
use the Union sample of SNe Ia to put constraints in cosmological parameters with and
without cosmic dust extinction in order to see how important is the influence of the dust
in the extracted values of the cosmological parameters. They use high-AB(z) and low-
AB(z) extinction models for ΛCDM and Quintessence DE models (hereafter QDE). For
the QDE model, in both extinction cases they find a ∼ 2% shift of Ωm to higher values
17
and a ∼ 2.7% shift of w to less negative values, compared to the parameters extracted in
a dust-free Universe. All these changes are within 0.4σ confidence levels (see the section
below about the statistical analysis method that we use). For the ΛCDM model they
find a shift of δΩm = 0.02 (7.2%) which corresponds to 0.55σ. Finally, Corasaniti (2006)
adopts a cosmic dust model with graphite grains with size of α = 0.1µm in order to study
the shift in the parameters using the Gold SNe Ia sample (Riess et al. 2004) and he finds
a shift of δw = 0.06 (6.25%) that corresponds to 0.38σ.
Except for the intergalactic medium dust, there is another source of extinction, the in-
terstellar medium dust inside the host galaxies of SNe Ia. The SNe Ia that are hosted by
early-type galaxies have a more tight light curve distribution than those hosted by late-
type galaxies (Riess et al. 1999; Sullivan et al. 2006) leading to a very useful uniformity.
This occurs due to the older star populations of early-type galaxies which have a smaller
range of masses than those in the late-type galaxies (Suzuki et al. 2012, hereafter S-12).
However, the SNe Ia colors do not seem to be affected significantly by the type of their
host galaxy.
As known, late-type galaxies contain notably more interstellar dust than early-type ones.
This means that the SNe Ia that are observed in the late galaxies, have smaller statistical
errors and several studies (Sullivan et al. 2003; 2010) have shown that these SNe Ia are
slightly better standard candles than those hosted by late-type galaxies. After correcting
for the color and light curve shape, the SNe Ia in early-type hosts seem to be about
0.14±0.07 mag brighter than those in later types (Hicken et al. 2009c; S-12). This factor
can lead to important systematics errors if not corrected. Finally, SNe Ia in early-type
hosts are more rare since, at lower-z it is observed that one in five SNe Ia occur in an
early-type. (S-12).
1.6 Other cosmological probes
Except for SNe Ia, there are many other probes to study the cosmological parameters,
like the temperature fluctuations of the CMB and Baryon Acoustic Oscillations.
18
Figure 3: The CMB power spectrum as measured by
WMAP until 2006
• CMB radiation is the picture
of our Universe when its age was
about 350, 000 years old, when
photon decoupling occured. It is
almost isotropic, with small tem-
perature fluctuations of ∆T ≈
10−5 K that are considered to
be one of the most powerful cos-
mological probes. The power
spectrum of CMB temperature
anisotropies versus their angular
scale δθ, or multipole moment l,
contains a wealth of cosmological information (Ryden, 2003) since its shape is deter-
mined by oscillations whose amplitudes and positions depend on the Universe’s compo-
sition. For example, the position of the highest peak of the ∆T versus δθ (or l) curve (at
l ≈ 200o, δθ ≈ 0.85o) determines the curvature of the Universe and is consistent with
a spatially flat geometry, k = 0. The characteristic angular scale of the first peak in
the CMB power spectrum is θ =rs(zls)
dA(zls)with rs(zls) is the comoving size of the sound
horizon at last scattering before the photon decoupling and dA(zls) is the comoving an-
gular distance to the last-scattering surface (Davis, Mortsell, Sollerman et al. 2007).
The anisotropies on larger angular scales derive from the influence of the gravitation of
primordial density fluctuations in the Dark Matter distribution (Ryden, 2003). The ratio
of heights between the first and second peaks shows us the amount of baryonic matter.
Generally, using the peaks of the spectrum and using advanced computer simulations to
provide the predictions of different models, we can put constraints in various parameters.
• The BAO provides a characteristic feature in the 2-p correlation function of galaxies,
which results from the interplay between the radiation pressure and gravity in the pho-
tobaryonic plasma, coupled rigidly due to the Thomson scattering in the Universe before
the recombination epoch (Basset, Hlozek, 2009). These two forces impose a system of
standing sound waves within the photon-baryon fluid (Wu et al. 2015). At recombina-
19
tion the pressure on the baryons sharply declines because of the capturing of the free
electrons by the atomic nuclei, and as a result the baryons have small over-densities at
the sound horizon. The corresponding scale is the distance the sound wave had travelled
in the plasma before the recombination (Wu et al. 2015). This distance can be measured
from the CMB to be about 147 Mpc and it can be also measured from the clustering
distribution of galaxies today. Therefore, the BAO scale is what we call a standard ruler.
Standard rulers are cosmological probes with a well known size at a redshift z , or with a
well known way of changing their sizes with redshift. So, by measuring the apparent di-
ameter of the object, we can find the angular distance dA(z) for the given z as well as the
luminosity distance dL(z), and using 1.7 we can constrain the cosmological parameters.
20
2 Theoretical expectations
In order to see how the distance modulus µ depends on the cosmological parameters
and the redshift z, we first compare models with different w and Ωm (we assume a flat
universe, i.e. Ωm + ΩΛ = 1) with a reference ΛCDM model with w = −1 and Ωm = 0.3.
Throughout this whole thesis we use H0 = 70 km/s/Mpc. We also compare models with
Ωk 6= 0 (non-flat Universe) with the same reference model and finally we use models with
the Chevallier-Polarski-Linder (CPL) parametrization of the equation of state parameter.
We define the relative deviations of the distance modulus as
∆µ = µmodel − µΛ (2.1)
where µ is given by eq. (1.6).
2.1 Different QDE and ΛCDM models comparison
To begin with, we first compare cosmological models with different Ωm with the reference
model and then with different w. We firstly use spatially flat models (Ωk = 0). For the
case of a constant w, the Hubble parameter, as we can see in eq. (1.5), is given by
H(z) = H0
√Ωm(1 + z)3 + Ωk(1 + z)2 + ΩΛ(1 + z)3(1+w) (2.2)
We use four different values for each parameter:
As we can see in Figure 4, the relative deviations of the distance moduli in the left panel
are much larger than those in the right panel. That indicates that it is easier to put
limits to the values of Ωm using SNe Ia with z ≥ 1 if we assume a certain value of w,
than limit the values of w for a certain Ωm. This happens because changes in Ωm affect
the luminosity distance (and hence the distance modulus) more strongly than changes in
w.
21
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1
∆µ
z
Ωm=0.15
Ωm=0.25
Ωm=0.35
Ωm=0.45
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.01 0.1 1
∆µ
z
w=-0.85
w=-0.95
w=-1.05
w=-1.15
Figure 4: The expected distance modulus deviations ∆µ vs. z Left panel: Between models
with w = −1 and different Ωm (= 0.15, 0.25, 0.35, 0.45) and the reference model (w = −1,
Ωm = 0.3). Right panel: Between models with Ωm = 0.3 and different w (=
−0.85,−0.95,−1.05,−1.15) and the reference model (w = −1, Ωm = 0.3).
In addition, we can see that in the left panel there is no maximum relative deviation of the
distance moduli. This happens because the first term under the square root containing
Ωm in eq. (2.2) is the most dominant of large redshifts. ∆µ increases with a lower rate
as z increases. The curvature of ∆µ changes at z ≈ 0.9− 1.
In the right panel, the maximum deviation from the reference model occurs at z ≈
1.3 − 1.4. This happens for the same reason as above. The first term under the square
root in H(z) is getting stronger as z increases while the two models have the same Ωm. As
a result, for z ≥ 1.4 any difference in H(z) between the two models begin to eliminate.The
curvature of ∆µ changes at z ≈ 0.5.
As can be derived from Friedmann equations and as shown here, lower values of Ωm and
w correspond to larger luminosity distances for the same z. This means that in a more
”empty” Universe, or in a Universe with a more negative equation of state parameter,
the radial velocity of any light source is lower for the same luminosity distance than in a
Universe with a higher value of Ωm or a less negative w.
22
2.2 Models comparison for a non-flat Universe (Ωk 6= 0)
In this subsection, we compare models with four different values for spatial curvature
density (Ωk = 1 − Ωm − ΩΛ) with a reference flat ΛCDM model. We use the 95%-
confidence limits for ΩK as derived by the CMB measurements.
In the left panel of Figure 5, we see that for negatively curved models, the distance
modulus µ is always larger than this of the reference model (∆µ > 0). The values of ΩK
are the 95%-confidence limits as gives by Planck Collaboration (2014). For the positively
curved models with Ωk = 0.006 and Ωk = 0.008 we have ∆µ < 0 until z = 3.09. For larger
redshifts we have ∆µ > 0 while their lowest peaks are both at z = 0.93. It is noteworthy
that for every model, negatively or positively curved, ∆µ tends to rise. The relative
magnitude deviations though are not large enough in order to exclude some values of Ωk
with our observations, except for the most negative one, Ωk = −0.086.
Specifically, for Ωk > 0, the maximum deviation is ∆µ = −0.003 for z=0.93 and ∆µ =
−0.0028 for z=5. This strong degeneracy means that we cannot distinguish models with
a very small Ωk > 0 from models with Ωk = 0. On the other hand, it is easier to test
models with a significant Ωk < 0 using our observations due to the relatively high ∆µ. An
interesting point is that the best fits given by all CMB surveys are provided for Ωk < 0.
2.3 Model comparison for CPL parametrization (w = w(z))
Next, we use the CPL parametrization of equation of state parameter:
w(z) = w0 + w1z
1 + z(2.3)
where w0 and w1 are constants. This form of w(z) indicates that the relation between
pressure P and density ρ, p = w(z)ρ, of the Dark Energy, changes with time. In this
case the Hubble parameter as defined in eq. (1.5) is given by
23
H(z) = H0
√Ωm(1 + z)3 + Ωk(1 + z)2 + ΩΛ(1 + z)3(1+w0+w1) exp
(− 3w1z
1 + z
)(2.4)
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.01 0.1 1
∆µ
z
Ωk=0.006
Ωk=-0.086
Ωk=0.008
Ωk=-0.029
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.01 0.1 1
∆µ
z
w0=-1, w1=0.3, Ωm=0.28
w0=-1.1, w1=0.3, Ωm=0.27
w0=-0.9, w1=0.1, Ωm=0.3
w0=-1.05, w1=0.2, Ωm=0.31
Figure 5: The expected distance modulus deviations ∆µ vs. z. Left panel: Between
models with (w = −1, Ωm = 0.3) and different Ωk and the reference model (Ωk = 0).
Ωk(= −0.086, 0.006) are the 95%-confidence limits from Planck+WMAP CMB results while
Ωk(= −0.029, 0.008) are the 95%-confidence limits from Planck+WMAP CMB results and
other cosmological probes. Right panel: Between spatially flat models with a combination of
w(z) and Ωm and the reference ΛCDM model (w = −1, Ωm = 0.3)
In the right panel we have four models with a time varying equation of state parameter,
w(z). For a lower w0(= −0.9) value with respect to the reference model and a low positive
w1(= 0.1) (blue-dotted line), it holds that ∆µ < 0 as expected. The lowest peak is again
at z ≈ 1.3, where Ωm starts to play the biggest role in H(z). On the other hand, for
w = −1.1 and for a relatively large w1 = 0.3 (green-dotted line), it holds that ∆µ > 0, but
the lower value of Ωm(= 0.27) has of course to do with this. Maybe the most important
conclusion from the comparison , is the strong degeneracy that exists between the two
other models (red and pink line). There is almost no difference until z = 0.5 and there is
a very small difference of ∆µ ≈ 0.04mag up to z = 5. This is typical of how difficult it is
to differentiate models with w = w(z) instead of a constant w even if Ωm is constrained
with some other cosmological probe. Finally, as we have mentioned several times before,
24
the distance modulus deviations that is caused by different w are reduced as the redshift
increases above some z ∼ 1.5. That causes models with similar Ωm’s to have ∆µ ≈ 0 at
very high z’s (which however is beyond any possibility of obtaining observable data).
25
3 The SNe Ia data
Union2.1 is the most recent compilation of SNe Ia of Supernova Cosmology Project(SCP).
It is the largest one so far, as it first consisted of 833 SNe which were drawn from 19
different datasets. After the lightcurve quality cuts, 580 of them remained. All usability
cuts are developed regardless of cosmological models.
3.1 Criteria of confirmation of SNe Ia type and quality cuts
Only one light curve filter was used for the fitting of Union2.1’s data, the Spectral
Adaptive Light curve Template (SALT2-1). SALT2-1 uses the whole set in order to
calibrate empirical light curve parameters when, at the same time, it typically assumes a
ΛCDM or QDE model. Unfortunately, a model assumption is necessary since it handles
SNe Ia in distances which are larger than those required by the linear Hubble’s law.
However it has been verified that this does not bias the cosmological parameters fit.
As it stated at (S-12) these light curve parameters are
• a colour parameter c, which contains the intrinsic SN colour and reddening due to
dust in its host galaxy (deviation of the mean B − V color of the SN Ia)
• an overall normalization to the time dependent spectral energy distribution (SED)
of a SN Ia, χ0
• the deviation of the average light curve shape, χ1 (or s = χ1 + 1)
The distance modulus is then:
µ = mmaxB −M + αχ1 − βc
with mmaxB the rest-frame peak magnitude in the B-band (P. Smale, 2010) and M the
absolute magnitude in the same band with c = χ1 = 0. Also, SALT2 and SALT2-1 use
α = 0.135 and β = 3.19. The initial light curve standardization results in the best-fitting
values for the time t0 of maximum in the B-band light curves. SNe Ia with observation
time tobs significantly different (several days) from t0 were excluded. In addition, SN Ia
26
candidates which have deviations from the best fit light curve model that are significantly
larger than their uncertainties are rejected.
Some of the high-z SNe that were discovered by the Hubble Space Telescope (HST)
Cluster Supernova Survey, were excluded because they do not have enough information
in their light curves for their type to be determined unambiguously (S-12). SNe that are
SNe Ia, are classified as secure, probable or plausible. A secure SN Ia has a spectrum that
directly confirms that is a SN Ia. Any other SN has to satisfy two conditions: Firstly, its
host galaxy should have spectroscopic, photometric and morphological properties consis-
tent with those of an early-type galaxy with no detectable signs of recent star formation
(S-12), and secondly its light curve shape should be more consistent with that of a SN Ia
and inconsistent with any other known SN types. A probable SN Ia is one that does not
have a secure spectrum but it satisfies one of the above two non-spectroscopic conditions
that are required for a secure classification. Finally, a plausible SN Ia is one that has
an indicative light curve but has not enough information to reject being of other types
(S-12). The Union2.1 further contains 12 secure SNe Ia and 2 probable ones with re-
spect to the Union2, while there are two SNe Ia, one secure (SCP06U4) and one probable
(SCP06K18) that are not included in the Union2.1 sample.
3.2 Redshift distribution of the Union2.1 sample
In this analysis we only use SNe Ia with z ≥ 0.02 in order to avoid uncertainties in
their estimated distances due to the local bulk flow. Thus, we only use the 546 SNe Ia
that have z ≥ 0.02. Union2.1 contains 23 SNe Ia more than the Union2 sample, 21 of
them with z ≥ 0.02 and 10 of them with z > 1. These 10 SNe Ia were discovered by
the Hubble Space Telescope (HST) Cluster Supernova Survey, as well as another 4 SNe
Ia with z ∈ [0.623, 0.973]. Of these 14 SNe Ia, 6 are members of galaxy clusters and
8 are members of non-cluster galaxies. High-redshift SN Ia are extremely important in
calculating the cosmological parameters, because at these redshifts, the differences of the
various DE models are more significant, as we saw in the previous chapter.
From Figure 6 we see that the number of the observed SNe Ia decreases, as expected,
27
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Fre
quen
cy
z
546 SN Ia events
Figure 6: Redshift distribution of the 546 SNe Ia with z ≥ 0.02 of the Union2.1 sample.
with increasing z. Half of these 546 SNe Ia are for z < 0.316. There are 141 SNe Ia
(25.82% of the sample we use) with z ∈ [0.02, 0.1] while the 141 most distant ones have
z ∈ [0.562, 1.414]. This distribution indicates the need for more high-redshift observa-
tions. Most of the data with z < 0.4 were taken from the Sloan Digital Sky Survey
(SDSS) while most data with z ∈ [0.4, 1] were taken by the SuperNova Legacy Survey
(SNLS). For z > 1, data were taken from the HST survey.
3.3 Mapping the Union2.1 sample
The Union2.1 catalogue contains the name of every SN Ia, its redshift z, the distance
modulus µ, the corresponding 1σ statistical error, σstat, of µ and the probability of each SN
Ia to be a member of a low mass galaxy. It does not contain the equatorial coordinates
(right ascension α and declination δ) of SNe Ia, unlike Union2. However, in order to
calculate the cosmological parameters for different sky regions, we needed the coordinates
of the SNe Ia. Therefore, we took the equatorial coordinates from the Union2 catalogue
for the common SNe Ia with Union2.1, which are 556 (1 SN Ia is excluded from Union2 )
28
and we found the coordinates of the rest SNe Ia from the following websites:
• http://supernova.lbl.gov/2009ClusterSurvey/cands/
• http://www.cbat.eps.harvard.edu/lists/Supernovae.html
• http://astro.berkeley.edu/bait/public html
Knowing the equatorial coordinates of all the Union2.1 SNe Ia, we can produce several
maps of the sample which can help us to visualize its angular distribution. First we make
the polar diagram of all data, with the redshift z being the radius and the right ascension
α being the angle.
1.4
1
0.6
0.2
0.2
0.6
1
1.4
4h
8h
0h
12h
20h
16h
060120180240300
Union2.1 data
Figure 7: Union2.1 data (580 SN Ia events) polar plot for redshift z and right ascension α
As we can see in Figure 7, the higher-z SNe Ia were detected preferentially along various
directions, some of which contain more data than others. For example, there are 97 SNe
Ia for α ∈ [2h, 3h], but there are only 26 SNe Ia for α ∈ [15h, 20h], 22 of them with
z < 0.15.
A more useful representation of the sample’s data is to produce the isosurface plot of the
SNe Ia for α and δ (declination). Doing this, we will be able to see in which regions of
the celestial sphere we have the most data.
29
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
80
90
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Declination
δ (degrees)
Right Ascension α (hours)
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
80
90
-180 0 180
Declination
δ (degrees)
Right Ascension α (degrees)
Figure 8: Upper panel: Union2.1 data (580 SN Ia events) plot for equatorial coordinates α
and δ. Bottom panel:The isosurface plot of the sample for α and δ
It is obvious from Figure 7 and Figure 8 that the data distribution is not uniform, since
most SNe Ia have
δ ∈ [−10, 10]. In addition, a lot of them have α ∈ [0h, 4h] ∪ [20h, 24h]1.This happens
because the observational campaigns cover very small solid angles of the sky (cluster of
galaxies etc.), so they detect SNe Ia in tight regions instead of searching the whole sky,
a fact which is dictated by the observational strategies necessary to detect SNe.
1From now on we are going to denote this region as α ∈ [20h, 4h]
30
Specifically, there are 263 SNe Ia (45.3% of the sample) within the above δ limits. For
a more narrow sky region with α ∈ [20h, 4h] and δ ∈ [−1.25o, 1.25o] (300 deg2, 0.463%
of the celestial sphere), the SDSS camera was used on the SDSS 2.5m telescope at the
Apache Point Observatory (APO), in order to search for SNe Ia. The observations were
made in the northern fall seasons of 2005 to 2007 (Betoule et al. 2014). That means that
in these limits there are plenty of data with intermediate z’s, as SDSS-II survey targeted
SNe Ia with z ∈ [0.05, 0.4].
We can produce the same plot for the galactic coordinates as well. In order to do that, we
have to convert the equatorial coordinates α and δ into galactic longitude l and galactic
latitude b. For this purpose we use the transformation equations:
l = 33o + arctan
(sin δ sin 62o.6 + cos δ sin (α− 282o.25) cos 62o.6
cos δ cos (α− 282o.25)
)
b = arcsin [sin δ cos 62o.6− cos δ sin (α− 282o.25)]
(3.1)
where all coordinates are in degrees (if α = xh then α = (15x)o).
In Figure 9 we can see that all the data with δ ∈ [−3, 3] from the previous graph, now
form a characteristic ”smile” in the southern galactic hemisphere. There are more data
in the southern galactic hemisphere (351 SNe Ia) than in the northern (229 SNe Ia). As
we mentioned above, we are going to use only the data with z ≥ 0.02, to avoid the local
bulk flow-based uncertainties. With this in mind, we are left with 335 SNe Ia south and
211 SNe Ia north. There are some ”data-rich” areas in the south hemisphere, unlike the
north hemisphere where the observations are more scattered and random.
There are no observations of SNe Ia for b ∈ [−5, 5] due to the Milky way’s plane. The
large amount of dust and gas absorbs the emitted radiation in the optical band, not
allowing us to observe possible SNe Ia.
31
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
80
90
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
Galactic latitude b (degrees)
Galactic longtitude l (degrees)
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
70
80
90
0 180 360
Galactic latitude b (degrees)
Galactic longitude l (degrees)
Figure 9: Upper panel: Union2.1 data (580 SN Ia events) plot for galactic coordinates l and
b. Bottom panel: The isosurface of the sample for l and b.
3.4 Statistical uncertainties of the distance moduli
In every analysis of data that are derived from an experiment or observation in any field of
Physics, possible errors or uncertainties of the measurements can affect our conclusions.
In order to have a full perspective of the statistical uncertainties of the distance moduli
µ of the SNe Ia, we present their uncertainty distribution.
In Figure 22, we see that the 1/3 of the data (173 SNe Ia) have a statistical uncertainty
32
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1
Freq
uenc
y
σstat
σstat for 546 SN Ia events
Figure 10: Statistical uncertainty distribution of the 546 SNe Ia with z ≥ 0.02 of the Union2.1
sample.
of σstat ∈ [0.15, 0.2] while 3/4 of the data (414 SNe Ia) have σstat ∈ [0.1, 0.25]. The mean
value of it is σstat = 0.2229. To compare these statistical errors with their corresponding
distance moduli, we calculate the ratioσstatµ
%. There are only 46 SNe Ia withσstatµ≥ 1%,
and 4 of them haveσstatµ≥ 2%. On the contrary, there are 294 SNe Ia (53.8% of the
sample) withσstatµ≤ 0.5%, 7 of them with
σstatµ≤ 0.25%. Thus, we see that the distance
modulus uncertainties for the Union2.1 sample are minor for the most SNe Ia but there
are exceptions for which the measurement of µ is not very trustworthy.
33
4 Cosmological parameters fitting
4.1 One-parameter models (Ωm or w)
In order to fit our data with a model and extract the best values of the model parameters,
we use the χ2 minimization analysis. We change step-by-step our model parameters values
within some limits and see which combination of them give
χ2 = χ2min. Then, we estimate the uncertainty of these parameters, based on a desired
confidence level. For the whole data analysis part of this thesis, we accept a flat cosmology
and therefore ΩΛ = 1− Ωm.
As a first step in our analysis, we are going to use one free model parameter. To this end,
we initially take p ≡ (Ωm) for a constant w = −1 (ΛCDM) and secondly we use p ≡ (w)
for a constant Ωm = 0.3 (QDE). We use the distance moduli of 546 SNe I (z ≥ 0.02) and
as the error in eq. (A.4), we take the statistical uncertainty, σstat that is provided by the
Union2.1 for each SN Ia.
Since we have one fitted parameter, Nf = 1, our degrees of freedom will be d.o.f. =
546− 1 = 545. From Table 5 we see that for Nf = 1 we have the 1σ (68.3%) confidence
level for ∆χ ≤ 1. Also, the 3σ confidence level is provided for ∆χ ≤ 9. The limits
within which we search for the best-fit values are Ωm ∈ [0, 0.65] for the density matter
and w ∈ [−2,−0.5] for the DE equation of state parameter. Our typical step sizes are
0.001 for Ωm and 0.002 for w.
Our results, as shown in Figure 11, are:
• For w = −1 (fixed):
Ωm = 0.278+0.013−0.014, → χ2
min/d.o.f. = 520.479/545 = 0.955
• For Ωm = 0.3 (fixed):
w = −1.056± 0.032, → χ2min/d.o.f. = 520.593/545 = 0.955
34
1
3
5
7
9
0.239 0.264 0.278 0.291 0.319
χ2-
χ2
min
Ωm
∆χ2 for w=-1
1
3
5
7
9
-1.155 -1.088 -1.056 -1.024 -0.959
χ2-
χ2
min
w
∆χ2 for Ωm=0.3
Figure 11: Left panel: ∆χ2 for Ωm values if w = −1 (fixed). Right panel: ∆χ2 for w if
Ωm = 0.3 (fixed)
We see that our fit in both cases is excellent, with χ2/d.o.f. = 0.955. That shows that
the best pair of values (w,Ωm) is very close to these values. Though, it is more efficient
to fit both parameters at the same time, as we do later. For the 3σ (99.73%) confidence
level (∆χ2 ≤ 9), we have Ωm ∈ [0.239, 0.319] and w ∈ [−1.154,−0.96], limits however
that require the second parameter to be fixed a priori.
4.2 One-parameter model (q0)
We now turn to study the deceleration parameter q0. For low to medium redshifts, we
can use an approximate to eq.(1.8) relation between luminosity distance dL and q0,
dL =c
H0
[z +
1
2(1− q0)z2 + . . .
](4.1)
Using eq.(4.1) and eq.(1.6) we can again apply the χ2 minimization analysis and calculate
q0. Since q is a function of time, q = q(t), we have to use redshifts so as q = q0, i.e. the
parameter value in our epoch. If we use redshifts, for example, z > 0.4, q will not be
approximately constant and our calculations would be wrong. We get the SNe Ia with
35
an increasing zmax with a step size of 0.02, in a range of z ∈ [0.1, 0.36]. For every zmax
we calculate dL and µ for q0 ∈ [−1.5, 0.5] with a step size of 0.001.
-0.9
-0.85
-0.8
-0.75
-0.7
-0.65
-0.6
-0.55
-0.5
-0.45
-0.4
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36
qo
zmax
qo versus zmax with 1σ error bars
Figure 12: Deceleration parameter q0 with 1σ error bars versus the maximum redshift z
until which we take data
As we can see when we plot our results in Figure 12, there are small differences for
different maximum z’s. As we increase zmax, the 1σ error bars get smaller, because
we have more data. For zmax = 0.1, we have 141 SNe Ia and q0 = −0.351+0.249−0.252 with
χ2/d.o.f. = 126.434/140 = 0.903. These large uncertainties (±71% of the best-fit value)
show that we cannot accept a ”safe” mean value of q0. On the other hand, for zmax = 0.24
we have 218 SNe Ia and q0 = −0.501 ± 0.086 with χ2/d.o.f. = 204.738/217 = 0.943. In
this case the 1σ uncertainty is much smaller than the previous (±17%). This is due to
the larger number of data. In both cases, χ2/d.o.f. is very satisfying. It is noteworthy
that none of the uncertainties limits pass to a positive value of q0. Every negative value
of q0 corresponds to a positive acceleration of the scale factor, α > 0. This indicates that
accelerated expansion of our Universe is a robust result even when using limited in z SNe
Ia data.
If we accept that q0 = −0.501 ± 0.086, from the Friedmann equation’s (1.1) and their
derivatives, eq. (1.2) and eq. (1.4), for a flat Universe (k = 0) we have:
36
q0 =Ωm
2− ΩΛ = −0.501⇒
Ωm
2− (1− Ωm) = −0.501± 0.086⇒
Ωm = 0.333± 0.057
(4.2)
The listed uncertainty has been calculated using error propagation, with δΩm =2
3δq0 and
δqo = 0.086. Thus, even if we only use the SNe Ia with z ≤ 0.24, we have a very good
approximation of the matter density Ωm , consistent with the full QDE model results
described below.
4.3 Two-parameter model (Ωm and w)
The main subject of this thesis is to calculate the best-fit values of the QDE cosmological
models, i.e. where p ≡ (w,Ωm), using the reduced χ2 analysis for a group of SNe Ia
data that we choose each time. We also want to plot the solution space within the
1σ uncertainty of our fit and see the limits of the parameters. We use only statistical
uncertainties, σ = σstat, as given in Union2.1 sample. We accept a QDE model of our
Universe.
First, we use the entire sample (546 SNe Ia with z ≥ 0.02) in order to constrain the two
free parameters of the model.
The procedure is as follows: For every w , we use values of the matter density Ωm ∈
[0, 0.65] with a step size of 0.002. When all values of Ωm are tested for a certain w, we
increase w by a step of 0.0025 in a range of w ∈ [−2,−0.5]. Thus, we have 325× 600 =
195, 000 different pairs of p ≡ (w,Ωm) to find the best one. For two fitted parameters,
Nf = 2, the 1σ error corresponds to ∆χ2 ≤ 2.3 while the 3σ error corresponds to
∆χ2 ≤ 11.83. We produce a contour plot that shows the best-fit values and the area
covered in the w and Ωm plane for ∆χ2 ≤ 2.3 and ∆χ2 ≤ 11.83.
In Figure 13 we show the two contours that correspond to the 1σ and 3σ confidence
levels. The corresponding ranges of the 1σ region are: Ωm ∈ [0.17, 0.364] and w ∈
[−1.2425,−0.8]. Moreover, the limits of 99.73% confidence level are: Ωm ∈ [0, 0.448]
37
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.2425
-1.005
-0.8
0.17 0.28 0.364
Figure 13: Solution space for w and Ωm as it occurs from the analysis of the whole sample
(546 SNe Ia). The blue contour corresponds to the 1σ confidence level (68.3%) and the grey to
the 3σ confidence level (99.73%). The limits of the 1σ confidence level as well as the best-fit
values are shown with the dashed lines.
and w ∈ [−1.61,−0.59]. These values may seem to have huge uncertainties but is very
important to understand the significance of knowing that there is a possibility of 99.73%
for the matter density to be Ωm ≤ 0.448 and the DE equation of state parameter to be
w ≥ −1.61. Of course, not every value of one parameter can be combined with any other
value of the second parameter in 1σ limits. For more negative values of w , Ωm increases
and vice versa.
Table 1: QDE cosmology fits for w and Ωm using all the SNe Ia data (546). When p ≡
(w,Ωm), the uncertainty of each parameter is given by the range for which ∆χ2 ≤ 2.3 while the
other parameter is fixed at its central (best) value.
w Ωm χ2min/d.o.f.
−1.005± 0.045 0.28+0.02−0.018 520.478/544 = 0.956
−1 (fixed) 0.278+0.013−0.014 520.479/545 = 0.955
−1.056± 0.032 0.3 (fixed) 520.593/545 = 0.955
38
The results presented in Table 1 show that our fit is very good, with a reduced χ2min ≈ 1.
The extracted values are very common for nowadays cosmology. Furthermore, we see
that more data mean lower uncertainties, since for the Constitution set (366 SNe Ia with
z ≥ 0.02) the uncertainties are ±0.053 for w and ±0.022 for Ωm (Plionis et al. 2011)
We can visualize our fit results by plotting the distance moduli µ of the SNe Ia versus their
redshifts z and using as error bars the σstat. In the same plot we show the theoretically
expected µ versus z for our best-fit model, for a matter dominated, Einstein-de Sitter
universe (w = 0,Ωm = 1) and, for a DE dominated, de Sitter Universe (w = −1,ΩΛ = 1).
35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Union2.1 data (546 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0 35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Union2.1 data in groups of five (110 groups)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0
Figure 14: Distance moduli µ as a function of redshift and the expected µ for three cosmo-
logical models. Left panel: For all SNe Ia (546). Right panel: For 108 groups with 5 SNe Ia
each and 2 groups with 3 SNe Ia each (the six most distant).
In the left panel of Figure 14 it can be easily seen that our best-fit model describes the
data way better than the Einstein-de Sitter model or the matter-empty model. The
Einstein-de Sitter model predicts that distant SNe Ia should be brighter by ≈ 0.6 mag
at z ≈ 1 and ≈ 0.7 mag at z ≈ 1.3 than our best-fit QDE model. Even at z ≈ 0.19 the
deviation of the two models is ≈ 0.2 mag. So, the model that was thought to describe
the Universe until 1998 is way off from the observed µ − z relation. In addition, Figure
14 clearly shows the need of observations of distant SNe Ia in order to obtain the best-fit
model with more accuracy.
39
In the right panel of Figure 14, we group the SNe Ia into groups aggregates in order the
tendencies to be more clearly visible. This is realized by sorting the data in ascending
redshift order and then group every 5 SNe Ia together, starting from the one with the
smallest redshift. The group’s µ, z and σstat are the mean values of the respective quan-
tities of the 5 members of the group. For the last two groups we reduce their membership
into 3 SNe Ia per group in order to have a slightly larger dynamical range in z.
4.4 Hubble flow divided in different redshift bins
In order to test whether the solution space of our parameters change, we dissociate our
data into three bins. The first one has 272 SNe Ia for z ∈ [0.02, 0.314]. The second and
third ones have 137 SNe Ia each and their limits of redshift are z ∈ (0.314, 0.57] and
z ∈ (0.57, 1.414] respectively. We make the space solution for all three bins in order to
find the best-fit values and the 1σ and 3σ confidence levels.
Figure 15 contains some very interesting results. It is obvious that the 1σ contour for
z ∈ (0.57, 1.414] has the lowest uncertainty compared to the two others. This happens
due to the fact that high-z SNe Ia are far more important in constraining the cosmolog-
ical parameters than low-z SNe Ia because the largest model deviations occur at high
redshifts, as it is shown in Figure 4. It is interesting to note that although the lowest-z
subsample has twice as many SNe Ia as the highest-z subsample, and thus one expects
lower random errors, the importance of the high-z regime in constraining the cosmolog-
ical parameters, as described in Section 3, is the factor that provides the most stringent
constraints seen in Figure 15.
However, the size of these contours with respect to the whole sample, are much larger
since they contain 1/4 of the data for z ∈ (0.314, 0.57] and z ∈ (0.57, 1.414] and rela-
tively low redshifts for the first z -range. The 1σ contour range of the fitted parameter
values for the three z -bins are: w ∈ [−2,−0.625], Ωm ∈ [0, 0.578] for the first shell,
w ∈ [−2,−0.615], Ωm ∈ [0, 0.54] for the second shell and w ∈ [−1.665,−0.5575], Ωm ∈
[0, 0.424] for the third shell.
40
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-0.6825
0.002
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.695
0.472
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.665
-0.9225
0.248 0.424
Figure 15: Solution spaces for w and Ωm The blue contours correspond to the 1σ confidence
level (68.3%) and the grey to the 3σ confidence level (99.73%). Upper left panel: For the 272
SNe Ia with z ∈ [0.02, 0.314]. Upper right panel: For the 137 SNe Ia with z ∈ (0.314, 0.57].
Bottom center panel: For the 137 SNe Ia with z ∈ (0.57, 1.414].
Another interesting fact that is presented in Figure 15 is the variable verticality of the
three contour plots. As the redshift increases, the shape of the contours becomes more
vertical with respect to the Ωm-axes and tend to be more parallel to the w-axes. This
will be more easily visible in the upcoming sections (Figure 24).
In Table 2 we show the best-fit results and their χ2min values for the three different z -bins.
For the first two bins, our best-fit values of w and Ωm seem out of the norm. These
huge differences between them and the best-fit values from the whole sample should be
attributed to the inability to successfully constrain the cosmological parameters when a
41
Table 2: Results of the fits for w and Ωm using the three redshift’s bins SNe Ia data (272,
137 and 137). When p ≡ (w,Ωm), the uncertainty of each parameter is given by the range for
which ∆χ2 ≤ 2.3 when the other parameter is fixed at its central (best) value.
w Ωm χ2min/d.o.f.
First bin, z ∈ [0.02,0.314]
−0.6825+0.055−0.0725 0.002+0.054
−0.002 255.352/270 = 0.946
−1 (fixed) 0.275+0.054−0.052 255.592/271 = 0.943
−1.042± 0.04 0.3 (fixed) 255.639/271 = 0.943
Second bin, z ∈ (0.314,0.57]
−1.695+0.165−0.1775 0.472+0.028
−0.026 117.806/135 = 0.873
−1 (fixed) 0.277+0.038−0.037 118.07/136 = 0.868
−1.052+0.082−0.084 0.3 (fixed) 117.99/136 = 0.868
Third bin, z ∈ (0.57,1.414]
−0.9225+0.06−0.0625 0.248± 0.026 146.761/135 = 1.087
−1 (fixed) 0.279± 0.026 146.81/136 = 1.079
−1.06+0.052−0.054 0.3 (fixed) 146.905/136 = 1.08
sample is dominated by low-z , which lead to a huge degeneracy between models. As we
explained in the previous sections, all the pairs of p ≡ (w,Ωm) within the 1σ contour
plots are of almost equal possibility.
For example, for z = 0.3, the deviation in distance modulus for two models (w,Ωm) =
(−1.005, 0.28) and (w,Ωm) = (−0.6826, 0.002), is ∆µ = 0.007 and for z = 0.5 it becomes
∆µ = 0.034. So, for small samples and low redshifts, models are almost indistinguishable,
42
thus all the combinations of w and Ωm inside the 1σ contour are equally possible.
In order to test how much the SNe Ia with intermediate redshifts affect the final space
solution, we produce the 1σ and 3σ contour plots without the medium redshift bin which
has z ∈ (0.314, 0.57] and contains 137 SNe Ia and we compare it with the solution space
that occurs from the analysis of the whole sample.
w
Ωm
∆χ2<11.83 for the 1
st and 3
rd redshift bins
∆χ2<2.3 for the 1
st and 3
rd redshift bins
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.2375
-0.985
-0.76
0.148 0.272 0.36
w
Ωm
∆χ2<2.3 for the 1
st and 3
rd redshift bins
∆χ2<2.3 for the Union2.1 sample
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.005
0.28
Figure 16: Solution spaces for w and Ωm . Left panel: The blue contour corresponds to the
1σ confidence level (68.3%) and the grey to the 3σ confidence level (99.73%) for the 409 SNe
Ia of the 1st and 3rd redshift bins. Right panel: 1σ contours for the whole Union2.1 sample
(red contour) and for the 409 SNe Ia from the 1st and 3rd redshift bins.
As we can see in Figure 16, in this case the resulting solution space is almost identical
with that of the whole sample. It causes just a small shift of the best-fit parameter values
(2.86% for Ωm , 2% for w ) and slightly larger 1σ contours of the parameters. Specifically
we can compare the two solutions below:
• For the joint 1st and 3rd redshift bins:
w = −0.985+0.0525−0.055 , Ωm = 0.272± 0.022 with
χ2min/d.o.f. = 402.398/407 = 0.989 with 1σ corresponding ranges, w ∈ [−1.2375,−0.76],
Ωm ∈ [0.148, 0.36]
43
• For the whole sample:
w = −1.005± 0.045 , Ωm = 0.28+0.02−0.018 with
χ2min/d.o.f. = 520.478/544 = 0.956 with 1σ corresponding ranges, w ∈ [−1.2425,−0.8],
Ωm ∈ [0.17, 0.364]
This probably implies that it is more efficient for future surveys to focus on observing only
low and relatively high redshift standard candles while omitting intermediate redshifts.
44
5 Investigating possible Hubble expansion anisotropies
5.1 Among the two galactic hemispheres
In order to test if there is any anisotropy in expansion between the two galactic hemi-
spheres, we perform the same work as above for each hemisphere separately. First present
the redshift distribution of their data.
0
5
10
15
20
25
30
35
40
45
50
55
60
65
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Fre
qu
en
cy
z
335 SN Ia events
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Fre
qu
en
cy
z
211 SN Ia events
Figure 17: Redshift distribution of the SNe Ia with z ≥ 0.02. Left panel: Southern galactic
hemisphere (b < 0, 335 SN Ia events) Right panel: Northern galactic hemisphere (b > 0, 211
SN Ia events).
Figure 17 shows us that the redshift distributions of the two galactic hemispheres are quite
distinct. The data of the southern galactic hemisphere are more normally distributed
while in the northern galactic hemisphere there are 82 SNe Ia (39% of the total) with
z < 0.1 and only 9 SNe Ia with 0.1 ≤ z < 0.3. This can make the differentiation of
the fitting cosmological models even harder in the northern galactic hemisphere. This is
because for low redshifts, distance moduli deviations ∆µ between the models are very
small to distinguish and much smaller than the statistical uncertainties. For example, at
z = 0.3 for two models (w,Ωm) = (−0.6, 0) and (w,Ωm) = (−2, 0.6) we have ∆µ = 0.01.
On the other hand, the two galactic hemispheres have similar number of SNe Ia for
z ≥ 0.4, 119 for the southern and 106 for the northern.
45
We follow the same procedure as for the whole sample, in order to calculate the best-fit
values for p ≡ (w,Ωm) for each galactic hemispheres. The grids,the contour plots and
the typical step size are the same with the whole sample’s analysis.
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.3175
-0.985
-0.725
0.108 0.274 0.394
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.5325
-1.07
-0.7225
0.12 0.3 0.416
Figure 18: Solution spaces for w and Ωm . The blue contours correspond to the 1σ confidence
level (68.3%) and the grey to the 3σ confidence level (99.73%). Left panel: For the 335 SNe Ia
of the southern galactic hemisphere. Right panel: For the 211 SNe Ia of the northern galactic
hemisphere.
In Figure 18 the corresponding ranges of 1σ contour for the southern galactic hemi-
sphere are Ωm ∈ [0.108, 0.394], w ∈ [−1.3175,−0.725] and for the northern are Ωm ∈
[0.12, 0.416], w ∈ [−1.5325,−0.7225]. These constraints of parameters, as expected, are
somewhat larger than the constraints from the whole sample. As the amount of data
increases, the uncertainties decrease. The 1σ contour of the whole sample is a subset of
the respective contours of the two hemispheres.
In Table 3 we present the results of the χ2min analysis for the southern and northern
hemisphere.
It can be seen that the best-fit values for the two hemispheres are slightly different but
within the 1σ uncertainty contours range. Both hemispheres’ solution spaces contain the
best-fit parameter values of the whole sample, as seen in Table 1. The value of χ2min/d.o.f.
is very close to unity, especially for the southern galactic hemisphere.
46
Table 3: Results of fits for w and Ωm using independently the southern and northern galactic
hemispheres’ SNe Ia data (335 and 211 respectively). When p ≡ (w,Ωm), the uncertainty of
each parameter is the range for which ∆χ2 ≤ 2.3 when the other parameter is fixed at its central
(best) value.
w Ωm χ2min/d.o.f.
Southern galactic hemisphere
−0.985± 0.0575 0.274+0.026−0.028 334.588/333 = 1.005
−1 (fixed) 0.28± 0.02 334.594/334 = 1.002
−1.0425± 0.04 0.3 (fixed) 334.664/334 = 1.002
Northern galactic hemisphere
−1.07± 0.08 0.3+0.03−0.028 185.741/209 = 0.889
−1 (fixed) 0.274± 0.02 185.817/210 = 0.885
−1.07+0.052−0.054 0.3 (fixed) 185.741/210 = 0.884
As we did for the whole sample, we plot the distance moduli µ with the statistical
uncertainties σstat versus the redshift of the SNe Ia for each hemisphere as well as the
best-fit models. Moreover, we plot the theoretical distance modulus expectations for the
Einstein-de Sitter and the de Sitter model.
In Figure 19 we see that the best-fit model for each hemisphere predicts almost the same
µ with the best-fit model of the whole sample for z < 1.5. In addition, the lack of data for
0.1 ≤ z < 0.3 in the northern galactic hemisphere is characteristic. Finally, the northern
data contain more ”outliers”, since the SNe Ia in redshifts z = 0.375, 0.55, 0.592 deviate
from the best-fit model by 3.98%, 4.25% and 3.33% respectively. Also, the first two have
the two largest statistical errors of the whole sample (σstat = 0.9232 and σstat = 1.006
respectively) and the third one has the fourth largest error (σstat = 0.718). The third
largest error of the sample belongs to the SN Ia at z = 0.97 of the northern galactic
47
35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Southern galactic hemisphere data (335 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-0.985, Ωm=0.274, ΩΛ=0.726
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0 35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µz
Northern galactic hemisphere data (211 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-1.07, Ωm=0.3, ΩΛ=0.7
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0
Figure 19: SNe Ia distance moduli µ as a function of redshift and the expected µ for four
cosmological models. Left panel: The southern galactic hemisphere’s 335 SNe Ia Right panel:
The northern galactic hemisphere’s 211 SNe Ia.
hemisphere again, which deviates from the best-fit model by 2.8%.
5.2 Among random groups of SNe Ia
In order to test further if there are any anisotropies in the Hubble expansion, we select
four solid angles that contain SNe Ia spanning a large dynamical range in redshift. Two
of them are in the southern galactic hemisphere and the other two in the northern. These
groups are:
• Group A with l ∈ [160, 185] and b ∈ [−63,−50] which contains 82 SNe Ia
• Group B with l ∈ [50, 80] and b ∈ [−70,−38] which contains 53 SNe Ia
• Group C with l ∈ [160, 220] and b ∈ [10, 60] which contains 46 SNe Ia
• Group D with l ∈ [85, 140] and b ∈ [45, 70] which contains 63 SNe Ia
In Figure 20 we present the results of the usual χ2-minimization procedure. We have
used the same step in Ωm and w as before.
48
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.605
-0.9125
0.21 0.426
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-2
-0.875
0.234 0.566
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.5175
-0.65
0 0.418
w
Ωm
∆χ2<11.83
∆χ2<2.3
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.785
-1.02
0.262 0.434
Figure 20: Solution space for w and Ωm . The blue contour corresponds to the 1σ confidence
level (68.3%) and the grey to the 3σ confidence level (99.73%). Upper left panel: As it occurs
from Group A (82 SNe Ia). Upper right panel: As it occurs from Group B (53 SNe Ia). Bottom
left panel: As it occurs from Group C (46 SNe Ia). Bottom right panel: As it occurs from Group
D (63 SNe Ia).
We see that for Group B and Group C the 1σ contours seem very different since they had
totally different best-fit values of w and Ωm . Although we have significantly fewer data
for each group than we had for the first redshift shell, z ∈ [0.02, 0.314], the uncertainties
in Group A and Group D are lower. This is a result of the variety of redshifts in these
groups as we see below. Specifically, the corresponding ranges of 1σ contours of these
groups are:
• Group A: w ∈ [−1.605,−0.5975], Ωm ∈ [0, 0.426]
49
• Group B: w ∈ [−2,−0.875], Ωm ∈ [0.234, 0.622]
• Group C: w ∈ [−1.5175,−0.555], Ωm ∈ [0, 0.418]
• Group D: w ∈ [−1.785,−0.5875], Ωm ∈ [0, 0.434]
Groups C which has the fewest data, also has the lowest 1σ corresponding range for w,
while its uncertainty for Ωm for every value of w is the highest.
In Table 4 we summarize the results of χ2 analysis for the four data Groups. It is
noteworthy that Groups A and D provide results very near to those of the whole SNe
Ia sample, and they also have the lowest χ2/d.o.f.. Of course, they have the most SNe
Ia compared with the other groups. Even more important is that they also have high-z
data unlike Groups B and C. There is one more interesting observation and is that for
Group B, χ2 for (w,Ωm) = (−1, 0.312) has an important deviation of the best-fit values,
∆χ2 = 1.88. There is no other deviation as large as this for w = −1 (fixed) from χ2min in
this thesis for any other group of data. This may be a hint for a small anisotropy in the
expansion of the Universe for these coordinates, something that we analyse in the next
chapter. In order to see the best-fit models with the corresponding data we produce the
µ vs.z plots for each group.
In Figure 21 we plot the SNe Ia distance moduli µ as a function of redshift z of each
Group. We also plot the best-fit model as it occurs for each Group, the best-fit model
for the whole sample (w,Ωm) = (−1.005, 0.28), the Einstein-de Sitter model and the de
Sitter model. It is evident that Groups A and D have several high-z SNe Ia (z ≥ 1),
while they lack in Groups B and C, with the highest redshift of Group B being z = 0.687
and of Group C being z = 0.882.
In addition, for Groups B and C, the value of χ2min/d.o.f. indicates that the fitted model
is not such a good fit to the data. The main characteristic of Group C is that it contains
the four SNe Ia with the largest uncertainties. The deviations of these distance moduli
∆µ from the theoretically expected µth are quite large. For example, the SN1997l with
z = 0.55 have a deviation of the best-fitted model of ∆µ = 1.795 mag with an uncertainty
of σµ = 1.006. Even for the de Sitter model (w = −1,ΩΛ = 1) it has ∆µ = 1.53 mag. If
50
Table 4: Complete table of fits for w and Ωm using Group A,B,C and D SNe Ia data (82,
53, 46 and 63 respectively). When p ≡ (w,Ωm), the uncertainty of each parameter is the range
for which ∆χ2 ≤ 2.3 when the other parameter is fixed at its central (best) value.
w Ωm χ2min/d.o.f.
Group A, l ∈ [160,185],b ∈ [−63,−50]
−0.9125+0.085−0.0875 0.21+0.048
−0.044 64.6/80 = 0.808
−1 (fixed) 0.253± 0.044 64.66/81 = 0.798
−1.11± 0.04 0.3 (fixed) 65.879/81 = 0.813
Group B, l ∈ [50,80],b ∈ [−70,−38]
−2+0.366 0.566± 0.054 60.02/51 = 1.177
−1 (fixed) 0.312+0.09−0.085 61.9/52 = 1.19
−0.996± 0.0168 0.3 (fixed) 61.938/52 = 1.214
Group C, l ∈ [160,220],b ∈ [10,60]
−0.65± 0.095 0+0.108 49.52/44 = 1.125
−1 (fixed) 0.248+0.088−0.079 50.3/45 = 1.12
−1.096+0.16−0.214 0.3 (fixed) 50.6/45 = 1.127
Group D, l ∈ [85,140],b ∈ [45,70]
−1.02± 0.095 0.262+0.042−0.039 54.779/61 = 0.898
−1 (fixed) 0.257+0.043−0.039 48.023/62 = 0.775
−1.116+0.118−0.126 0.3 (fixed) 48.09/62 = 0.776
51
35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Group A data (82 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-0.9125, Ωm=0.21, ΩΛ=0.79
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0 35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Group B data (53 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-2, Ωm=0.56, ΩΛ=0.44
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0
35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Group C data (46 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-0.65, Ωm=0, ΩΛ=1
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0 35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Group D data (63 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-1.02, Ωm=0.262, ΩΛ=0.738
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0
Figure 21: Distance moduli µ as a function of redshift and expected µ for four cosmological
models. Upper left panel: Group A (82 SNe Ia) Upper right panel: Group B (53 SNe Ia).
Bottom left panel: Group C (46 SNe Ia). Bottom right panel: Group D (63 SNe Ia).
we take into account the other ”outliers” of this Group as well we can explain why we
have such a high value of χ2min/d.o.f. for Group C. On the other hand, the data of Group
B do not seem to have ”outliers” and they generally have small uncertainties. But that
exactly is the reason of the high value of χ2min/d.o.f. and the apparently bad fit. The low
values of the uncertainties, being in the denominator of eq. (A.4), do not allow significant
deviations of the data from the model. This means that even for a very good fit to the
data, if two or three SNe Ia have a deviation of ∆µ ≈ σµ the χ2 value rises, resulting in
this case to a χ2min/d.o.f. = 1.177.
52
For the Groups A and D, the fit of the data are quite satisfying. Especially for Group D,
the best-fit values are very close to the best-fit model from the whole sample, (w,Ωm) =
(−1.005, 0.28). Furthermore, the deviations of distance moduli of the models for these
Groups compared to those from the whole sample, are very small even for z = 1.5 (for
Group D, ∆µ = 0.038). In contrast, the deviations ∆µ for the best-fit models to the
data of Groups B and C are larger at high redshifts (∆µ = −0.26 and ∆µ = 0.223 for
z = 1.5 respectively) but insignificant at the redshift range covered by the data. If there
is any anisotropy in the expansion of the Universe for these directions, it is not easily
seen, although the results based on Group B (l ∈ [50, 80] and b ∈ [−70,−38]) have the
most distinct 1σ contour in the Ωm -w place with respect to all the other Groups.
To find out if this peculiarity of the solution space of Group B is a result of the low-z
data, we also find the solution space for the data of Group A but for the same upper-z
limit (z ≤ 0.689). We also find the solution space for the whole Union2.1 sample after
limiting its redshift to the same upper-z limit (z ≤ 0.689).
w
Ωm
∆χ2<2.3 for the Group A data with z<0.689
∆χ2<2.3 for Group B
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.32
0.382
w
Ωm
∆χ2<2.3 for Group B
∆χ2<2.3 for Union2.1 data with z<0.689
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.12
0.334
Figure 22: 1σ contour plots of the Ωm -w plane. Left panel: As they occur from data from
Group A with z ≤ 0.689 (blue contour, 67 SNe Ia) and Group B (red contour, 53 SNe Ia). Right
panel: As they occur from Union2.1 with z ≤ 0.689 (blue contour, 453 SNe Ia) and Group B
(red contour, 53 SNe Ia).
53
In Figure 22 we see that if we use only the data with z ≤ 0.689, the best-fit values come
closer to those of Group B than before. For clarity, we use only the 1σ contour plots. For
Group A, we have (w,Ωm) = (−1.32, 0.382) with χ2min/d.o.f. = /65 = and for Union2.1
we have (w,Ωm) = (−1.12, 0.334) with χ2min/d.o.f. = 427.433/451 = 0.948. Although
there is a significant overlapping between the 1σ contours, especially for Group A, the
difference of the solution space for the same maximum redshift appears still important.
Thus, constraining the upper redshift limit brings the results based on Group B closer to
those of Group A and of the whole sample but still there is significant discrepancies.
54
6 Joint likelihood analysis
Once we have the solution spaces for several independent groups of SNe Ia, we can use
them to perform a joint likelihood analysis and compare the extracted results with those
from the χ2 analysis from the whole sample.
6.1 Joint likelihood analysis for the two galactic hemispheres
First we consider the southern and the northern galactic hemispheres’ distance moduli
as our two independent groups of data.
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
w
Ωm
∆χ2<11.83 for the northern galactic hemisphere
∆χ2<11.83 for the southern galactic hemisphere
∆χ2<2.3 for the northern galactic hemisphere
∆χ2<2.3 for the southern galactic hemisphere
w
Ωm
∆χ2<11.83 for the joint analysis of the galactic hemispheres ∆χ
2<2.3 for the joint analysis of the galactic hemispheres
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.2425
-1.005
-0.8
0.17 0.28 0.364
Figure 23: Left panel: 1σ and 3σ contour plots of southern (blue and grey contours) and
northern (red and green contours) galactic hemispheres. Right panel: Solution space for w and
Ωm as it occurs from the joint likelihood analysis for the two galactic hemispheres. The blue
contour corresponds to the 1σ confidence level (68.3%) and the grey to the 3σ confidence level
(99.73%)
As we can see in the left panel of Figure 23, the contours of the southern galactic hemi-
sphere are smaller and less vertical with respect to the Ωm-axes than the correspond-
ing northern ones. The joint analysis of the two hemispheres provides χ2min/d.o.f. =
520.478/(333 + 209) = 0.96 for (w,Ωm) = (−1.005, 0.28). In the right panel of Figure 23
we have the solution space as it occurs from the joint analysis. As expected, the best-fit
55
values, the corresponding ranges of 1σ and 3σ contours as well as the value of χ2min are
exactly the same with these from the full sample analysis. The only difference is that the
d.o.f. are reduced and the χ2min/d.o.f. is closer to unity than before.
6.2 Joint likelihood analysis for the three redshift bins
The next subdivisions of the sample that we are going to use are the three redshift bins
as they defined before. The first bin contains 272 SNe Ia for z ∈ [0.02, 0.314], the second
contains 137 SNe Ia for z ∈ (0.314, 0.57] and finally the third also contains 137 SNe Ia,
for z ∈ (0.57, 1.414].
w
Ωm
∆χ2<2.3 for the first redshift shell (0.02<z<0.314)
∆χ2<2.3 for the second redshift shell (0.314<z<0.57)
∆χ2<11.83 for the third redshift shell (z>0.57)
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
w
Ωm
∆χ2<11.83 for the joint analysis of the 3 redshift shells ∆χ
2<2.3 for the joint analysis of the 3 redshift shells
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.2475
-1.01
-0.8025
0.172 0.282 0.366
Figure 24: Left panel: 1σ contour plots of the three redshift bins (for clarity we do not
plot any other contours). Right panel: Solution space for w and Ωm as it occurs from the
joint likelihood analysis for the three redshift bins. The blue contour corresponds to the 1σ
confidence level (68.3%) and the grey to the 3σ confidence level (99.73%)
From this joint analysis, the results are slightly different from those from the whole sample
due to round-up errors. The values with the lowest χ2 are (w,Ωm) = (−1.01, 0.282)
with χ2min/d.o.f. = 520.735/(270 + 135 + 135) = 0.964. The best-fits values and the
uncertainties of them when one parameter is fixed are the same with these from the
whole sample. The best-fit combination for the whole sample (w,Ωm) = (−1.005, 0.28)
56
have a deviation of ∆χ2 = 6.6× 10−4, being almost an equally good fit that is extracted
by the joint analysis. Indeed, it is the combination with the lowest ∆χ2 of all.
As we mentioned before for higher redshifts the corresponding contours are more vertical
to the Ωm-axes. Also, we see more clearly that the third and more distant redshift bin
provides the smallest size contour region, although has half the data of the first bin. This
is an excellent indication of the importance of the high-z data.
6.3 Joint likelihood analysis of ten independent SNe Ia subsam-
ples
In order to identify possible anisotropies of the Hubble expansion, we now divide the
Union2.1 sample in 10 fully independent SNe Ia subsamples. Except for the Groups
A,B,C and D we also have:
• Group E with l ∈ [245, 360] and b ∈ [10, 80] with 56 SNe Ia
• Group F with l ∈ [90, 158] and b ∈ [−70,−50] with 68 SNe Ia
• Group G with l ∈ [25, 80] and b ∈ [−35, 90] with 47 SNe Ia
• Group H with l ∈ [81, 215] and b ∈ [−50,−5] which contains 54 SNe Ia
• Group K with l ∈ [30, 45, ] and b ∈ [−90,−40] with 22 SNe Ia
• Group M with all the remaining 55 SNe Ia
We do not use letters ”I, J, L” because they can be mistaken with ”1” and with each
other. Group K has the fewest data but it has important, high-z data with a mean
redshift of z = 0.6847.
In section 5 we pointed out the peculiarity of the solution space provided by Group B.
Since Group K is very near to Group B in galactic coordinates (Figure 9), we join the
two Groups (B and K) in one group and calculate again the solution space of w and
ΩmThe area of these groups in celestial coordinates is between α ∈ [21.2h, 23.94h] and
δ ∈ [−26.36o, 13.46o]. We have to mention here that Group K alone gives similar best-fit
57
values (w = −2,Ωm = 0.51) with Group B and the fact that it has the closest coordinates
to Group B is a strong hint for a consistent behaviour of this whole region.
The best-fit values of the cosmological parameters provided by the joint B+K subsample
are w = −1.7025+0.25−0.275 and Ωm = 0.498+0.042
−0.04 while the 1σ ranges are: w ∈ [−2,−0.8825]
and Ω ∈ [0.262, 0.574]. However due to the degeneracy of the solution, the best-fit
parameters of the whole SNe Ia sample are not included within the 1σ contour provided
by the B+K group.
When we exclude every other Group from the whole sample, the solution spaces does not
change significantly. This means that the 1σ contours for all Groups except for B and
K, are consistent with the results of the rest of the data.
58
7 Further analysis for the unusual sky region
Since we identified this sky region with a distinctly different (w -Ωm ) solution with respect
to all other regions of the sky and of the whole sample together, we now slightly expand
the subsample to include all SNe Ia in this part of the sky, which is limited by the
following galactic coordinates:
35o < l < 83o
−79o < b < −37o
This sky region, which we call Group X, contains 82 SNe Ia (7 more than Groups B+K).
Performing the usual χ2min minimization procedure, we find the best-fit values for Group
X and plot the 1σ contour plot, to be compared with the 1σ contour plot of the rest of
the SNe Ia.
w
Ωm
∆χ2<2.3 for Group X
∆χ2<2.3 for the rest of the data
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-0.955
0.25
w
Ωm
∆χ2<2.3 for Union2.1 sample without Group X
∆χ2<2.3 for the whole Union2.1
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.005
0.28
Figure 25: Left panel: 1σ contour plots of Group X (red contour) and for the rest of the data
(blue contour). Right panel: 1σ contour plots for the whole Union2.1 sample (red contour)
and for the sample without Group X (blue contour).
• For Group X:
w = −1.74+0.245−0.26 , Ωm = 0.504+0.04
−0.038 with
χ2min/d.o.f. = 91.203/80 = 1.14 with 1σ corresponding ranges, w ∈ [−2,−0.94] ,
59
Ωm ∈ [0.288, 0.574]
• For the rest of the data:
w = −0.955± 0.045 , Ωm = 0.25+0.022−0.02 with
χ2min/d.o.f. = 425.528/462 = 0.92 with 1σ corresponding ranges, w ∈ [−1.195,−0.7475],
Ωm ∈ [0.126, 0.342]
As we see in the left panel of Figure 25 the 1σ contour plot of the Group X data is
significantly different than that of the rest of the sample with the two contours having
nearly no common solution space. In the right panel, there is a 5.2% change in the best-fit
w value and a 12% change in the best fit Ωm value when we use all the data compared
with the case where we exclude Group X from the analysis. The shift in the Ωm best-fit
value is quite important since it indicates that the amount of matter in the Universe,
may be overestimated because of the behaviour of Group X SNe Ia.
35
36
37
38
39
40
41
42
43
44
45
46
47
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
µ
z
Group X data (82 SN Ia events)
w=-1.005, Ωm=0.28, ΩΛ=0.72
w=-1.74, Ωm=0.504, ΩΛ=0.496
w=-1, Ωm=0, ΩΛ=1
w=0, Ωm=1, ΩΛ=0
Figure 26: Distance moduli µ of Group X (82 SNe Ia) as a function of redshift and expected
µ for four cosmological models.
This statistically important difference between Group X and the rest of the SNe Ia may
be due to an anisotropy of the Hubble expansion in this area of the sky or due to some
systematic observational error. In order to further investigate these possibilities, we
60
search the SNe Ia individual surveys that constitute the SNe Ia of Group X. We referred
to the site;
http://astro.berkeley.edu/bait/public html
and we find that, from the 82 SNe Ia of Group X, there are:
• 36 SNe Ia from the Sloan Digital Sky Survey II (SDSS-II, 43.8% of the data) with
redshift range z ∈ [0.043, 0.387]
• 19 SNe Ia from the Supernovae Legacy Survey (SNLS, 23.2% of the data) with
redshift range z ∈ [0.285, 0.961]
• 18 SNe Ia from the Equation of State: SupErNovae trace Cosmic Expansion survey
(ESSENCE, 22% of the data) with redshift range z ∈ [0.205, 0.687]
• 9 SNe Ia from several other surveys (2 from LOSS, 2 from HST, 1 from ESCC and
4 from individuals, 11% of the data) with redshift range z ∈ [0.024, 1.192]
A large amount of the Group’s data was observed by the SDSS-II, so this survey has
a dominant role to any results extracted from Group X. In addition, as we mentioned
before, the SDSS-II contains SNe Ia with z ≤ 0.4. ESSENCE and SNLS also play a key
role to the results since each survey provide about 1/4 of the 82 SNe Ia and both have,
especially SNLS, quite large redshifts.
To see how much the different sky surveys affect the final χ2 minimization results, we
perform the Jackknife Resampling method. Accordingly, we analyse the data of Group
X excluding a certain subsample each time. To this end, we calculate the best-fit models
and the 1σ contour plots for Group X, firstly without the SDSS-II data, secondly without
the SNLS data and finally without the ESSENCE data. The results are:
• Group X without SDSS-II data (46 SNe Ia):
w = −0.7875+0.1175−0.1225 , Ωm = 0.256+0.076
−0.07 with
χ2min/d.o.f. = 47.637/ = 1.083
61
w
Ωm
∆χ2<2.3 for Group X without SDSS-II data
∆χ2<2.3 for Group X
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-0.7875
0.256
w
Ωm
∆χ2<2.3 for Group X
∆χ2<2.3 for Group X without SNLS data
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-2
0.566
w
Ωm
∆χ2<2.3 for Group X
∆χ2<2.3 for Group X without ESSENCE data
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-1.83
0.496
Figure 27: 1σ contour plots for Group X (red contour) and for Upper left panel: Group
X without the SDSS-II data (blue contour), Upper right panel: Group X without the SNLS
data (blue contour), Bottom center panel: Group X (red contour) and for Group X without the
ESSENCE data (blue contour).
• Group X without SNLS data (63 SNe Ia):
w = −2+0.345 , Ωm = 0.566+0.052−0.048 with
χ2min/d.o.f. = 66.029/61 = 1.0824
62
• Group X without ESSENCE data (64 SNe Ia):
w = −1.83+0.2675−0.17 , Ωm = 0.496+0.042
−0.038 with
χ2min/d.o.f. = 76.767/62 = 1.238
The upper left panel of Figure 27 shows that excluding the SDSS-II data from the Group
X, we obtain a dramatic shift of the 1σ contour with respect to that of Group X unlike
the upper right and the bottom center panel, in which we have sequentially eliminated
the SNLS and the ESSENCE data respectively. Without the SDSS-II data, the ”banana-
shaped” 1σ contour shifts to a more consistent position with respect to that of the whole
SNe Ia. Moreover, excluding the SNLS or ESSENCE data, the SDSS-II subset becomes
more dominant in Group X, causing the 1σ contour to shift into even more negative values
of w and larger values of Ωm . The conclusion that can be drawn from Figure 27 and from
the above χ2min values is that this unusual behaviour of this sky region, is determined by
the SNe Ia of the SDSS-II survey. In other words, responsible for this difference in the
w -Ωm solution behaviour between the sky region of Group X and the rest of the sky are
the SDSS-II data in this region and not an actual anisotropy in the expansion. If there
was an anisotropy in Hubble flow for this sky region, it should probably be seen also at
all distant SNe Ia as well.
To see if there is a general systematic error in the SDSS-II survey, we compare the
solution results of many other SDSS-II SNe Ia from the sky region where the SDSS-
II focused from 2005 to 2007 (excluding Group X SNe Ia) as we mentioned in Section
3, with those of Group X. We remind the reader that this region is this characteristic
”smile” in the southern galactic hemisphere or, in equatorial coordinates, α ∈ [20h, 4h]
and δ ∈ [−1.25o, 1.25o]. We set an upper limit in redshift, 0.02 < z ≤ 0.4 since this is the
maximum redshift of SDSS-II data. Under these circumstances, 115 SNe Ia are left that
we can use within these coordinates and redshift limits. Obviously, SNe Ia from other
surveys can be included in these subsample but the vast majority of data comes from the
SDSS-II, allowing us to extract conclusions regarding a possible systematic behaviour of
the SDSS-II.
63
w
Ωm
∆χ2<11.83 for the SDSS-II data
∆χ2<2.3 for the SDSS-II data
-2
-1.7
-1.4
-1.1
-0.8
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6
-0.6825
0.026
Figure 28: Solution space for w and Ωm as it occurs from the analysis of the SDSS-II
region without the data from Group X (115 SNe Ia). The blue contour corresponds to the 1σ
confidence level (68.3%) and the grey to the 3σ confidence level (99.73%). The limits of the 1σ
confidence level as well as the best-fit values are shown with the dashed lines.
The results are seen in Figure 28 where the best-fit parameter values are w = −0.6825±
0.0625 and Ωm = 0.026+0.078−0.026 with χ2
min/d.o.f. = 97.734/113 = 0.865. Thus, all the other
SDSS-II data provide a normal solution space behaviour compared with the SDSS-II from
Group X. This probably means that the different behaviour of the Group X SNe Ia is not
a result of some SDSS-II systematic error nor of an anisotropy in the Hubble expansion.
We suggest as a possible solution of this unusual behaviour a large bulk flow attending the
redshifts of the SNe Ia of Group X with z ∈ [0.043, 0.387] . We are currently investigating
this possibility in detail
64
8 Conclusions
Several conclusions can be drawn based on the results of this thesis. First of all, high-z
SNe Ia are necessary in order to distinguish the QDE cosmological models. Observa-
tionally, there are sky regions with far more SNe Ia than others, mainly in the southern
galactic hemisphere. The best-fit cosmological parameters values with their 1σ uncertain-
ties for the Union2.1 sample, if we use the QDE cosmology model, are Ωm = 0.28+0.02−0.018,
w = −1.005±0.045 and q0 = −0.501±0.086, which is consistent with the other published
results and the ”concordance” cosmological model (Suzuki et al. 2012). We have searched
for possible anisotropies of the Hubble expansion, by dividing the sample into different
solid angles. We found no significant anisotropy in the Hubble expansion between the
two galactic hemispheres. However, we identified a particular sky region, with galactic
coordinates 35o < l < 83o & −79o < b < −37o, which provides a statistically significant
different w − Ωm solution with respect to the rest of the sample or any other subsample
that we have analysed. Although this could have been attributed to an anisotropy in
the Hubble flow, we found that this explanation is not supported by our analysis but
rather this result could well be due to a large bulk flow of galaxies at this region within
z ∼ 0.15− 0.3.
65
Appendices
A χ2 minimization, our data analysis method
The most common statistical method to see how well a theoretical model describes ob-
served data, is probably the χ2 (chi-squared) minimization procedure. Below we describe
the basics of this method which will be used throughout this thesis.
Suppose that we make a measurement (xi, yi) of a quantity y for a given xi. The expected
value for yi according to the theory would be f(xi) and the total error of the experimental
measurement (due to uncertainties of the measurement, etc.) would be σ. The probability
to make this measurement is 1
P (xi) =1
σi√
2πexp
[−1
2
(yi − f(xi)
σi
)2]
(A.1)
Now suppose that we make N measurements of this quantity that are independence of
each other. The total probability of obtaining this entire set of these N data points is
equal to the product of the probability of each data point (C. Laub, T. Kuhl), so
Ptot =N∏i=1
P (xi) =
[N∏i=1
1
σi√
2π
]exp
[−1
2
N∑i=1
(yi − f(xi)
σi
)2]
(A.2)
If we want to find the maximum probability we have to minimize the sum in the expo-
nential term of Ptot, so, we define this quantity as
χ2 =N∑i=1
(yi − f(xi)
σi
)2
(A.3)
and for χ2min we have the maximum probability. Thus, if the theoretical function f(xi,p),
depends on a set of parameters, which correspond to the elements of the vector p, we
can test for which values of these parameters we get the maximum probability.
1Probabilities are normalized to their maximum values (Plionis et al. 2011)
66
In our case, yi is the observed distance modulus of the each SN Ia, µobs(zi), which comes
with its redshift zi. Furthermore, f(xi) is the theoretical expected distance modulus
µth(zi,p) for the same redshift zi, with p ≡ (w,Ωm), for example. So, we have
χ2 =N∑i=1
(µobs(zi)− µth(zi,p)
σi
)2
(A.4)
with σi being the error in the distance modulus µi of each SN Ia. As we said before,
we have χ2min for the best-fitting parameters p0. When the parameters differ from these
values, p 6= p0, then χ2 increases, so ∆χ2 = χ2 − χ2min > 0. Limits of ∆χ2 that depend
on the number of the fitted parameters Nf , define confidence regions that contain a
certain fraction of the probability distribution of p’s 1. Almost every possibility region
corresponds to a certain standard deviation σ. All these are shown in Table 5.
Table 5: ∆χ2 limits for Nf fitted parameters and different confidence regions
σ P Nf = 1 Nf = 2 Nf = 3
1σ 68.3% 1 2.3 3.53
90% 2.71 4.61 6.25
2σ 95.4% 4 6.17 8.02
3σ 99.73% 9 11.83 14.2
In order to see how good our fit is, we have to divide the minimum χ2min by the degrees of
freedom, d.o.f. and get the reduced chi-square, χ2min/d.o.f. If we have N data points and
Nf fitted parameters, then d.o.f. = N −Nf . For a satisfactory fit we want χ2min/d.o.f. ≈
1. In other words, our model describes the observational data very well, inside our
uncertainties limits. If χ2min/d.o.f. 1 it means that our model is not an appropriate
fitting function or that we have not taken into account some form of unknown systematic
error. Finally, we may have underestimated our observational uncertainties (Bremer,
2009). On the other hand, if χ2min/d.o.f. 1, probably we have overestimated the
1For more information see ”Numerical Recipes 3rd Edition: The Art of Scientific Computing”, William
H. Press
67
statistical uncertainties and systematics. As a result, the 1σ possibility will cover a very
wide range of the fitted parameters, wide range of parameters, a non-desired effect.
B Joint likelihood analysis
In joint likelihood analysis, we take n fully independent data samples. For every combi-
nation of (w,Ωm) (we have a total of 195,000 combinations), we obtain a joint χ2 which
is given by
χ2 = χ21 + χ2
2 + · · ·+ χ2n (B.1)
while the degrees of freedom become
d.o.f. = d.o.f.1 + d.o.f.2 + · · ·+ d.o.f.n (B.2)
where (χ21 . . . χ
2n) are the values of χ2 for every one of the n independent data samples
while (d.o.f.1 . . . d.o.f.n) are the degrees of freedom of the χ2 analysis for every sample.
We then find the minimum, χ2min and the deviation of all the others from it, ∆χ2 =
χ2 − χ2min. Then, we plot the 1σ and 3σ contours (∆χ2 ≤ 2.3 and ∆χ2 ≤ 11.83
respectively) as they occur from the joint analysis.
68
9 References
Aguirre, A. N. 1999a, ApJ, 512, L19
Aguirre, A., & Haiman, Z. 2000, ApJ, 532, 28
Amanullah, R., et al. 2010, ApJ, 716, 712
Basset, B.A. & Hlozek, R. 2009, arxiv.org/abs/0910.5224
Betoule M. et al., 2014, ArXiv e-prints
Bianchi, S., Ferrara, A., 2006, MNRAS, 358, 379 ApJ, 596, L135
Corasaniti, P.S., 2006, MNRAS 372,191
Davies, J. I., Alton, P. B., Bianchi, S., Trewhella, M., 1998, MNRAS, 300, 1006
Davis, Mortsell, Sollerman and ESSENCE, 2007, astro-ph/0701510
Guy, J., et al. 2007, A&A, 466, 11
Hamuy, M., Phillips, M. M., Suntzeff, N. B., Schommer, R. A., Maza, J., & Aviles, R. 1996,
AJ, 112, 2391
Hicken, M., et al. 2009c, ApJ, 700, 1097
Kim, A.G., et al. 1997, ApJ, 476, L63
Kowalski, M., et al. 2008, ApJ, 686, 749
Laub, C. & Kuhl, T. A Critical Look at the Fitting of Reflectivity Models using the Reduced
Chi-Square Statistic. Retrieved from http://neutrons.ornl.gov
Menard, B., Kilbinger, M., & Scranton, R. 2010, MNRAS, 406, 1815
Perlmutter, S., et al. 1998, Nature, 391, 51
Perlmutter, S., et al. 1999, ApJ, 517, 565
Planck Collaboration 2014 A&A 571, A16
Plionis M., Terlevich R., Basilakos S., Bresolin F., Terlevich E., Melnick J., Chavez R., 2011,
MNRAS, 416, 2981
Riess, A. G., et al. 1998, AJ, 116, 1009
69
Riess, A. G., et al. 1999, AJ, 117, 707
Riess, A. G., et al. 2004, ApJ, 607, 665
Riess, A. G., et al. 2007, ApJ, 659, 98
Ryden, Barbara. Introduction to Cosmology. Ohio State Uni., 2003
Varvoglis, H & Seiradakis, J.H.,: Introduction to modern Astronomy, Thessaloniki: Gargatanis
, 1994
Shustov, B. M., Vibe, D. Z., 1995, Astronomy Reports, 39, 579
Sullivan, M., et al. 2003, MNRAS, 340, 1057
Sullivan, M., et al. 2006, ApJ, 648, 868
Sullivan, M., et al. 2010, MNRAS, 406, 782
Suzuki N. et al., 2012, ApJ, 746, 85
Wu, P., Zhengxiang, L. & Hongwei, Y., 2015, arXiv:1501.01818
Yoon, S.C. & Langer, N., 2004, Astron.Astrophys.435:967-985,2005
70